Unenhanced CT-based predictive model to identify small bowel necrosis in patients with mechanical small bowel obstruction

Abstract

Objectives To investigate the diagnostic value of unenhanced CT in mechanical small bowel obstruction (SBO) with small bowel necrosis, and to establish a predictive model. Methods From May 2017 to December 2021, the patients with mechanical SBO admitted to our hospital were retrospectively collected. Taking pathology-confirmed small bowel necrosis as the gold standard, the experimental group was composed of patients with small bowel necrosis confirmed by pathology, and the control group was composed of patients with no intestinal necrosis confirmed by surgery or successful conservative treatment with no recurrence of intestinal obstruction during 1-month followed-up. Results A total of 182 patients were enrolled in this study, 157 patients underwent surgery, of which 35 patients were accompanied with small bowel necrosis and 122 patients were not (33 patients with ischemic findings at surgery without necrosis). Finally, there were 35 patients in the experimental group and 147 patients in the control group. Multivariable logistic regression showed that increased attenuation of small bowel wall (P = 0.002), diffuse mesenteric haziness (P = 0.010), difference of CT value between mesenteric vessel and aorta (P = 0.025) and U-/C-shaped small bowel loop (P = 0.010) were independent risk factors for the diagnosis of mechanical SBO with small bowel necrosis. Through internal verification, the area under curve (AUC) of the predictive model reached 0.886 (95%CI: 0.824–0.947), and the calibration result was moderate. Conclusion Multiple features (increased attenuation of small bowel wall; difference of CT values between mesenteric vessel and aorta; diffuse mesenteric haziness; and U-/C-shaped small bowel loop) of unenhanced CT have clinical value in the diagnosis of mechanical SBO with small bowel necrosis. The predictive model based on these four features could achieve satisfactory efficiency.